From a question to a testable investigation
Autorite series · 3/4 · Method. Scope, comparisons, predeclared decisions and traceable records.
A question may be clear while its test remains ambiguous. “Is history useful?” is an example: without specifying the history, task and comparison, any outcome can be read in several ways.
Autorite's method seeks to reduce that ambiguity through research contracts and traceable decisions.
1. Give the question a scope
The representation-sufficiency contract names the source, representation and inquiry or action family. Distinctions to preserve, permitted collapses, resource assumptions and excluded future inquiries must be stated.
Preservation requires a transformation, feature, tolerance and path structure. Reachability requires a source, targets, resources, rules and horizon. These contracts help readers determine whether similar claims answer the same question.
2. Admit competing explanations
Testing a mechanism only where it has an advantage cannot establish its necessity. RP002A separates conditions where the present suffices, history adds information, and variable history is irrelevant to the future.
It also compares different information access: present observation, finite history, learned recursive state and a reference knowing the generator law. Readers can ask whether an advantage came from access, learning procedure or budget.
3. Declare decision rules before outcomes
Measures and decision conditions are stated in advance. Which comparison seeks a benefit, and which tests equivalence within a declared tolerance?
Failing a benefit criterion does not automatically establish equivalence. A comparison may remain indeterminate under its predeclared rule. Replacing that label after seeing the data can change the research question.
4. Separate records and interpretation
Working code, completed execution, auditable outputs and an established scientific claim are different stages. Closing a package does not close the entire program.
Versions, data, weights, traces and deviations remain in the record. Replay can show consistency between saved outputs and the reported process without proving a general theory about the world.
Independent training replicates must also be distinguished from timepoints within a replicate. Many observations do not automatically create many independent experiments.
5. Publish the boundary
Assumptions, alternative explanations, excluded claims and new questions accompany the result. Negative and indeterminate outcomes remain. The next cycle can draw from the whole record rather than only its most attractive finding.
This method guarantees no success. Its narrower purpose is to make the answered question, resulting change and remaining uncertainty readable to others. The final chapter follows these steps through the completed RP002A study.
Continue the series
Sources and revision
Public Autorite revision inspected on 10 October 2026. Links are pinned to the same commit. Explanatory examples are editorial interpretations; reported findings retain their source scope.
This connects to
Chapter 3 of the Autorite series.
Explores the Autorite research program through this step of the series.